Environmental Injustice and Well Water Contamination in North Carolina
ENEC 698 Final Report
Spring 2017 Semester | May 2, 2017 Dr. Andrew George
Outline
I. Introduction
A. Background
B. Problem Statement
C. Research Question
II. Literature Review
III. Methods
IV. Analysis
A. Descriptive Statistics
B. Contaminant Co-occurrence
C. Sampling point
D. Age of Well (new vs. old)
E. Environmental Injustice Metrics (race, education, income) and Belews Creek
F. True Homes development
V. Conclusion
I. INTRODUCTION
Private wells play an essential role in supplying water to North Carolinians. It is
estimated that about a third of the state’s population, or over three million people, depend on
private wells for safe drinking water [1]. Private well owners are responsible for the testing and
safety of their well due to the fact that private wells are outside of the scope of the Safe Drinking
Water Act, which was originally passed in 1974 by the Environmental Protection Agency (EPA)
[2]. Fortunately for the state of North Carolina, the state’s general assembly enacted a statute
that, starting in 2008, required all newly constructed wells to be tested for contamination. These
tests are accessible to the public through the North Carolina Department of Health and Human
Services, as long as the owners of the wells submitted their test results to the local health
department.
With so many North Carolinians depending on private wells for their water needs, it is
imperative that the water quality meets contaminant level standards set out by the EPA, known
as maximum contaminant levels (MCL). Contaminated drinking water raises serious public
health concerns, as elevated levels of toxins such as lead or arsenic can be dangerous when
consumed in large quantities [3]. To investigate whether or not private wells were safe to drink
from, well water tests were analyzed from the following counties: Wayne, Stokes, Union, and
Robeson. These counties were chosen based on the presence or absence of coal ash pits, diverse
geographical locations across the state, and other factors.
This study aimed to examine whether Socioeconomic Status (SES), or Environmental
Justice metrics such as race, education, and income, predict private well contamination (natural
or anthropogenic), test frequency, and test protocols. Results of this study will provide insight
into whether private well contamination can demonstrate areas/factors of environmental
injustice.
II. LITERATURE REVIEW
In order to determine a research question that was both manageable in the relatively short
amount of time that a semester offered us and novel enough compared to past research, a
literature review was conducted. To begin, we examined the writings of Dr. Robert Bullard.
Considered to be the father of the environmental justice movement, Dr. Bullard investigated
various cases of environmental injustice. One of his books, Dumping in Dixie, investigated the
efforts of several African American communities to seek justice when confronted with
environmental problems that disproportionately impacted them, He found that race was the most
significant predictor of environmental injustice, even when socioeconomic class was considered
[3]. This finding inspired the inclusion of environmental justice metrics in our analysis. Based on
the idea that race and other factors like socioeconomic status, education level, and income could
potentially be predictors of environmental injustice, we decided to include them in our study.
Dr. Rebecca Fry’s research group in the UNC Gillings School of Public Health conducted
a study that was published in 2014 regarding heavy metal exposure through private wells and the
prevalence of birth defects in the study population in North Carolina. Arsenic, cadmium,
manganese, and lead were examined in this study because all of them are “known developmental
toxicants” which can cause developmental harm on an unborn baby if the mother is exposed
while pregnant [5]. The study was semi-ecologic, and used a cohort of babies (20,151) born
between 2003 and 2008. The North Carolina Birth Defects Monitoring Program monitored cases
based on 12 different birth defects that were considered to be important consequences of heavy
metal exposure according to the CDC. Census data was used to match the cases to the child’s
address at birth; this location was also compared to existing mapped information about the
location of wells throughout the state. Manganese was found to be associated with heart defects
in children. Arsenic was only mildly associated with defects, but there was a measured effect of
both arsenic and manganese present in well water. In the analysis described in the subsequent
sections, several of the approaches that were used by Saunders et al. were utilized. We included
these metals in our study based on the significant health effects they are known to have and their
inclusion in the Saunders study. Additionally, this study informed our use of census data and
some of the other mapping techniques that our group used.
In a study conducted by a team of researchers at Virginia Tech, the Flint water crisis was
examined through the lens of environmental justice. This research team included Dr. Kelsey
Pieper, who was instrumental in the early stages of our own analysis. This study examined the
role that weaknesses of public health policies and federal regulation of drinking water quality
play in perpetuating environmental injustice [4]. Specifically, the study examines how public
misunderstanding and inadequate lead monitoring can actually serve to undermine the original
purpose of the EPA’s Lead and Copper Rule. For the purposes of our analysis, this study
contained important background information about the shortcomings of federal regulation of
drinking water safety, which is likely carried over into state policies.
Dr. Pieper was the lead author of another study that examined water samples from the
“ground zero” home in Flint, Michigan. This was one of the studies that brought waterborne lead
poisoning to the forefront of national news and public attention. In this study, the researchers
deduce that the cause of the elevated lead levels in drinking water was a result of the
“destabilization of lead-bearing corrosion rust layers that accumulated over decades” within the
piping that fed into homes [10]. The analysis of the samples from this ground zero home
eventually lead to the declaration of a state of emergency in Flint, and likely prevented further
exposure to lead (although the problem is far from resolved). It should be noted that the water
crisis in Flint disproportionately impacted minority populations and areas known for lower
socioeconomic status. This is clearly an incident of environmental injustice, and further
emphasizes similar cases need to be investigated.
Miguel de Franca Doria devoted time to investigating how people perceive their water
quality. In a 2009 study, conducted in the UK and Portugal, Doria looks into the factors by which
surveyed people evaluate the quality of their drinking water and the rate at which surveyed
people use bottled water instead of tap water in the household. The methods of this study
consisted of focus groups and a survey. The survey, being the qualitative component of the
study, asked participants to rank their agreement with statements such as “My tap water is
usually of high quality,” “I trust my tap water company,” and “I am happy with the odor of my
tap water” [11]. These results were analyzed using both structural equation models and
generalised linear models, with results suggesting that “perceptions of water quality and risk
result from a complex interaction of diverse factors.” The most significant findings of this study
give that flavor and peer opinions are the most important factors in consumer evaluation of risk.
This is significant because many harmful metals, including arsenic and lead, can’t be detected by
taste, smell, or appearance. Without testing, it is very possible that contamination could go
undetected, resulting in negative health effects in households and communities.
Dr. Sarah Flanagan published an influential three-paper series discussing arsenic in
private well water. The first part discusses an investigation of arsenic contamination in relation
to socio-economic status, henceforth abbreviated as SES. Because of the prevalence of naturally
occurring arsenic due to geologic formation, findings suggest one cannot predict the
concentration of arsenic with SES. In the third part of this paper, though, Flanagan discusses the
ways SES and education level affect the detection of and response to contaminants in well water.
People of higher SES and with higher levels of education are much more likely to know how to
get their water tested and be able to afford the tests. They are also more able to respond to
detected contamination, finding and buying alternate sources of water in the meantime and
finding and affording the resources, such as filters and more intensive repairs, that would
alleviate the problem [12]. This, combined with the increased vulnerability of low-SES people
shows the importance of well water contamination as a social justice issue.
III. METHODS
The counties used in this study- Stokes, Wayne, Union, and Robeson -were chosen for
analysis based off location in order to obtain a geospatial distribution of the state of North
Carolina. Individual well tests were obtained through the North Carolina State Laboratory Public
Health database of inorganic chemistry, where all new wells since 2008 are required to be tested
and reported in the database[5]. Over the course of 6 weeks, well test results were selected from
this database based on the specified counties. From 3732 well permits, 2625 valid reports were
manually recorded.
The following data points were extracted from the well test results: name of system,
street address, well identification, date collectected, date received/report date, collected
by/reported by, sample type, sample source, sampling point, and results in mg/L for arsenic,
cadmium, chromium, lead, manganese, and mercury. When values were not present in the well
test results, values were recorded as “NA”, and contaminant values that were under the
detectable limit were recorded as “ND”. Initial manual recording for data points such as sample
type/sample point/sample source into the spreadsheet was not standardized, meaning the same
text provided in the well test result was entered into the spreadsheet. Finalization/ standardization
of the spreadsheet was completed during later statistical analysis. However, street addresses were
standardized in the spreadsheet to use in geospatial geocoding before statistical analysis.
Through the use of the geospatial analysis program Esri ArcMap, the locations of the
individual well test were geocoded and plotted using the geocoding tool. Geocoding requires a
streetline shapefile to match the list of addresses to the well test results spreadsheet to a streets
database provided by the streetline shapefile. This shapefile, the integrated statewide road
network (ISRN) was downloaded via the North Carolina Department of Transportation
website[6]. Additionally, county boundaries were downloaded from this website. Census data,
which was required to map environmental justice metrics such as race, income and educational
attainment were downloaded from the American Fact Finder, a online search engine for the
Census.gov website [7]. A shapefile for census block groups polygons were downloaded via the
Geospatial Data Gateway, an online geospatial portal by the USDA [8]. Using Arctools in
ArcMap, maps were generated relating census data and well contaminants in order to determine
any significance between geographical location, race, income, and education. Statistical
programs such as SPSS and R were also used in order to determine any significance between
sample point of collection, age of well, types of contaminants, and occurrence of contaminants.
IV. ANALYSIS
A. Descriptives
Initial analysis began with compiling the test results extracted from the 2625 valid reports.
Contaminant level was compared to the state-determined maximum contaminant level (MCL)
and by county:
Wayne Contaminant Number of
wells tested Maximum Average MCL Wells above MCL
Arsenic 214 0.005 0.005 0.01 2 Cadmium 214 0.001 0.001 0.005 1
Lead 214 1.6 0.117 0.015 13 Manganese 214 0.18 0.053 0.05 52
Mercury 214 0 0 0.002 0 Union
Contaminant Number of wells tested
Maximum Average MCL Wells above MCL
Arsenic 1716 0.172 0.024 0.01 309 Cadmium 1716 0.006 0.004 0.005 1
Lead 1716 0.48 0.022 0.015 34 Manganese 1716 5.9 0.357 0.05 545
Mercury 1716 0.0014 0 0.002 0 Robeson
Contaminant Number of wells tested
Maximum Average MCL Wells above MCL
Arsenic 75 0 0 0.01 0 Cadmium 75 0 0 0.005 0
Lead 75 0.038 0.0155 0.015 3 Manganese 75 0.08 0.0538 0.05 5
Mercury 49 0 0 0.002 0 Stokes
Contaminant Number of wells tested
Maximum Average MCL Wells above MCL
Arsenic 598 0.108 0.0114 0.01 7 Cadmium 598 0.0018 0.0001 0.005 0
Lead 598 0.111 0.0097 0.015 80
Manganese 598 1.5 0.131 0.05 294 Mercury 594 0 0 0.002 0
The contaminant with the most hits above the MCL was manganese, with 52 in Wayne County,
545 in Union County, 5 in Robeson, and 294 in Stokes. lead and arsenic were second to
manganese.
In Wayne and Robeson counties, the average contaminant level for both lead and manganese is
higher than the set MCLs. In Union County, lead, manganese and arsenic all have average
contaminant levels above the MCL. Stokes County has average contaminant levels of arsenic and
manganese higher than their respective MCLs.
B. Contaminant Co-occurrence
The following table lists the number of double occurrences throughout the 2625 analyzed
well tests. Note the table lists co-occurrence of contamination at any level, not above the MCL.
Contaminants # of wells
Arsenic & Cadmium 4
Arsenic & Lead 57
Arsenic & Manganese 214
Cadmium & Chromium 1
Cadmium & Lead 1
Cadmium & Manganese 8
Chromium & Lead 14
Chromium & Manganese 18
Chromium & Mercury 1
Lead & Manganese 124
Manganese & Mercury 1
The highest level of co-occurrence was seen with Arsenic and Manganese, at 214 out of 2625
well tests (8.15%). Further analysis would examine co-occurrence in reference to sampling point
and age of the well.
C. Sampling Point
Each county’s well test database listed a variety of choices for sampling point of the
water tested. Our initial database of 2625 well tests contained over 20 options for ‘sampling
point,’ which were reduced to 6 options: inside, outside, well, port, unspecified/no response, and
miscellaneous.
Sampling Pt. Arsenic Cadmium Chromium Lead Manganese Mercury
Inside 0.004071925
091 0.006 0.035 0.00231146030
2 0.1093898724 0
Misc. 0.005701492
537 0 0 0 0.07341985816 0
Outside 0.004436967
89 0 0 0.00206324789
7 0.1071192183 0
Port 0.005 0.001 0.015 0.00539125431
5 0.02498698413 0
Well 0.004034820
593 0 0.03921724138 0.00269008914
7 0.1047519838 0.0014 Unspecified/NA
0.004082626421 0.0005785 0.02252444444
0.002704187867 0.1055584769 0
Above is a table of average contaminant values compared to sampling point listed on the
well test. There is slight variation throughout each of the contaminants’ averages according to
sampling point, but no statistical significance was found amongst contaminant level and
sampling point.
D. Environmental Justice Metrics and Belews Creek
There were three environmental justice metrics that were explored in our analysis of
private well water contamination: race, education, and income. (1) The four sub-groups used to
qualify racial composition were non-hispanic white, non-hispanic black, American indian, and
hispanic/latino. (2) Education level was assessed in a binary fashion, so the only education
threshold considered was the achievement of a bachelor’s degree. (3) Lastly, private well
contamination was looked at in the context of income, with which block groups were divided
into those with populations above and below 65% of the median income in North Carolina
(approximately $31,000).
While racial composition was considered as an environmental justice indicator for each
of the four counties, the random distribution of contaminants in Wayne and Union counties and
limited sample size in Robeson county made it difficult to make any conclusions about race and
other environmental justice metrics. For example, in Union County, census block groups with
populations of 20-45% non-Hispanic Black had only 4.8% of all arsenic contamination in the
county, and 5.5% of manganese. Wayne county had a random distribution as well. Census block
groups with 28-99% of non-Hispanic Black residents, contained 9.6% of manganese and 2.4% of
lead contaminants in the county. Due to the small number of well results racial GIS analysis was
omitted for Robeson county. Education and Income GIS results were also omitted for Wayne,
Robeson, and Union counties.
Of the four considered, Stokes county displayed interesting findings with respect to using
environmental justice metrics as predictors of well-water contamination. For this reason, GIS
results for each environmental justice metric are included for Stokes county. Stokes is different
from the three other counties because there is a significant pattern of contaminant occurrence.
This deviation from the random well-water contamination distribution is seen in the southeastern
corner of the county, which is also the location of a coal ash pit belonging to Belews Creek
Power Station.
Manganese appears to naturally occur throughout Stokes county. For this reason, natural
co-occurrence of the metal is expected to and does occur occasionally throughout the county.
Without taking into account the block group containing Belews Creek, there is only one well
with a naturally occurring metal that is not manganese-- this metal being chromium. Furthermore
and still excluding the same block group, there is no example of triple co-occurrence of any three
metals. The map of the Belews Creek area above illustrates a surprisingly different distribution
from the rest of the county. While it does not happen in any other Stokes county block group,
co-occurrence of arsenic, lead, and manganese contamination is found in 18 different wells on
the county-facing side of the Belews Creek Power Station. As if this were not already a
significant difference from the rest of the county, the multiple cases of arsenic, cadmium, and
chromium in this block group--neither of which appearing to naturally occur in Stokes
county--makes the cluster of contaminated wells even more interesting.
Belews Creek presents the opportunity to explore several environmental justice metrics
as possible predictors of well contamination. If wells in Belews Creek are so different, can we
find population characteristics that are unique to that area in order to explore a causal
relationship between the two?
In the GIS map below, there is a significant pattern of racial composition of the Belews
Creek block group compared to the rest of the county. The block group has a 20-62%
non-hispanic black racial composition. In addition, 100% of arsenic and cadmium well
contaminants were located in this census block group. A problem with concluding any causality
here is the lack of gradient with the racial composition and contamination of other block groups.
While there is still significant racial variability throughout the county, the distribution of
contaminants like manganese and lead are still distributed randomly, regardless of the block
group.
When looking at education level, the cluster falls in the Belews Creek census block group
with less than two percent of its population having attained a bachelor’s degree, despite having
all cadmium and arsenic contaminants in the county. In order to prove any relationship between
well water contamination and education level, similar levels of contamination would need to be
found in other block groups with equally undereducated populations, but in the map below, the
block groups directly to the North of Belews Creek and farther to the West (both are filled with
white) do not show levels of contamination that are anywhere near what is seen in Belews Creek.
The final environmental justice metric considered was income. Using GIS, mapping of
contaminants was done as was seen previously, except instead of racial composition or education
level, block groups were characterized as being either above or below 65% of the median income
in North Carolina. Since Belews Creek is the only seemingly non-random cluster of
contaminants in the county, a true correlation between contamination and level of income would
suggest that block groups with similar income level should have a similar level of contamination.
Since this is not the case, it was concluded that income, like race and education, is not a
significant predictor of private well water contamination in Stokes county or any of the other
three counties.
One last analysis performed through ArcMap was the location of wells tested by the True
Homes contractors in relation to contaminants and income. As seen below, the areas where True
Homes is present are areas of higher income that do not fall below 65% of the median income.
From the combined True Homes well locations, 17% contained chromium, 6% contained lead,
5.8% contained manganese, and 1.2% contained arsenic. The idea that environmental injustice
can be detected based off income is negated because all of the True Home locations are above
the 65% of the median income threshold, but still contain some detection of contaminants within
the wells. This suggests it is possible wealthy communities may be placed on areas where
naturally occurring contaminants are present within the ground. These results further support
income is not a significant indicator of well water contamination.
V. CONCLUSION
In this project, we sought to evaluate the extent to which traditional environmental
injustice metrics, including minority prevalence and socioeconomic status of a region, could
predict the presence of contamination in private wells used for residential drinking water. To
investigate this relationship, we extracted data from the North Carolina State Laboratory Public
Health database of Environmental Inorganic Chemistry, which has well test data on private
wells, and compared it to information from the census about race and education across 4 North
Carolina counties: Stokes, Wayne, Union, and Robeson. Although we did not find support for a
link between environmental injustice predictors and well contamination, we acknowledge that
other studies have found evidence to the contrary. Though our study does not add to this body of
evidence for environmental injustice in North Carolina, we postulate that patterns of
contamination may be linked to the metrics herein or others in ways that could not be elucidated
in a study of this scale. For example, a lack of publicly available test data in Robeson County (75
tests in DHHS database) may limit our ability to observe patterns. Importantly, Robeson County
is 25th in North Carolina counties (USGS) for the number of people using well water, suggesting
that the majority of people who use well water are not having their wells tested by the state, if
they are being tested at all.
No clear patterns between racial composition and well results were found in any of the
four counties examined. However, this was difficult to evaluate given the fact that we only
mapped contaminated wells. Though the contaminated well results do not appear to cluster in
any meaningful way, it appears as though certain census tracts do have more hits than others,
possibly reflecting a greater reliance on wells. This in and of itself could constitute an
environmental injustice, if those who are served by presumably cleaner municipal water differ in
significant demographic ways from those relying on well water.
A similar possibility exists for the lack of patterns in education and income. However, as
seen previously in Union county, certain naturally occurring contaminants like arsenic may be
less likely to have a disproportionate impact on low SES regions. However, we cannot ignore the
fact that either synergistic or antagonistic effects may be occurring to some degree: in other
words, can race and SES together explain a greater amount of variability in well results, or do
they actually have a negative effect on each other, making it harder to tease apart the factors
contributing to contamination? Understanding the relative influence of these and other factors
will be paramount to evaluating to the extent to which environmental injustice may be occurring
in well water across North Carolina.
In our study, manganese was by far the most common contaminant. Given the relatively
small amount of attention that has thus far been given to manganese as a human health threat, we
emphasize the importance of continued studies on the impact of this metal on North Carolinians.
Arsenic was also found to be exceptionally common, despite well known negative health
impacts. Furthermore, we found these two contaminants together more frequently than any other
combination in the study (214 instances). Co-occurrence of contaminants was also a notable
problem in Belews Creek, near the Belews Creek Power Station. We believe this may be due to
the coal plant’s coal ash pit nearby, but we cannot be sure without additional studies, including
species tracking to distinguish the contaminants from natural versus industrial sources.
Our study was limited in scope primarily by the small pool of tests from which our data
was extracted. The state database contains only information on publicly tested wells, meaning no
data could be obtained from wells for which owners chose to hire private companies to test the
water. Additionally, wells constructed prior to 2008 are not covered by General Statute 87-97,
meaning owners are not required to test the wells for contaminants. This may be playing a
significant role in Robeson county in particular, as the DHHS databases was particularly lacking
in tests despite the county’s high reliance on well water. Wells in Robeson may also be generally
older than those in other counties, which are experiencing significant development. Age of wells
will be an important factor to consider in future studies, especially given the fact that older wells,
legally, need not be tested at all.
To fully understand the relationship between environmental injustice metrics and well
contamination in North Carolina, future studies must not only consider well age, but the temporal
variability that may exist over time for any given well. Geographic factors like regional
topography and well depth may also influence the presence of contaminants. Crucially,
evaluating these large-scale variations will require expanding beyond the four counties examined
herein. However, small-scale studies will also be needed to understand what is happening on a
meaningful level in places like Belews Creek, where industrial contamination may be driving
high prevalence of contamination, and the True Homes developments, where these factors may
be less relevant. Perhaps most importantly, future studies should consider ways to get at data
from privately tested wells. There is no way to be certain just how much data is missing, though
we can safely assume that private tests represent a reasonable fraction of the data collected in
North Carolina. No analysis will be complete without incorporating these results. Ultimately, the
work has only begun and we strongly advocate for increased attention to the contaminants
common to NC, the patterns of contamination that may exist, and the nuances of these patterns.
References [1] Barros, Nirmalla, Kenneth Rudo, and Mina Shehee. "Importance of Regular Testing of Private Drinking Water Systems in North Carolina." North Carolina Medical Journal 75.6 (2014): 429-34. Web. [2] Barros [3] Bullard, R. D. (1990). Chapter 2: Race, Class, and the Politics of Place. Dumping in Dixie (pp. 21–36). [4] Katner, Adrienne, Kelsey J. Pieper, Yanna Lambrinidou, Komal Brown, Chih-Yang Hu, Howard W. Mielke, and Marc A. Edwards. "Weaknesses in Federal Drinking Water Regulations and Public Health Policies That Impede Lead Poisoning Prevention and Environmental Justice." Environmental Justice 9.4 (2016): 109-17. Web. [5] Sanders, A. P., Desrosiers, T. A., Warren, J. L., Herring, A. H., Enright, D., Olshan, A. F., Meyer, R. E., et al. (2014). Association between arsenic, cadmium, manganese, and lead levels in private wells and birth defects prevalence in North Carolina: a semi-ecologic study. BMC Public Health. Retrieved from http://bmcpublichealth.biomedcentral.com/articles/10.1186/1471-2458-14-955 [6] North Carolina Public Health. (n.d.). North Carolina State Laboratory Public Health Environmental Sciences - Inorganic Chemistry. Retrieved from https://slphreporting.ncpublichealth.com/EnvironmentalSciences/InOrganicChemistry/Default.aspx [7] North Carolina Dept of Transportation. (n.d.). GIS Data Layers. Retrieved from https://connect.ncdot.gov/resources/gis/pages/gis-data-layers.aspx [8] United States Census Bureau. (n.d.). American Fact Finder. Retrieved from https://factfinder.census.gov/faces/nav/jsf/pages/searchresults.xhtml?refresh=t [9] United States Department of Agriculture- Natural Resources Conservation Service. (n.d.) Geospatial Data Gateway. Retrieved from https://datagateway.nrcs.usda.gov/GDGHome.aspx [10] Pieper, K., Tang, M., Edwards, M. A. (2017). Flint Water Crisis Caused By Interrupted Corrosion Control: Investigating “Ground Zero” Home. Environmental Science and Technology. [11] Doria, Miguel De Franca, Nick Pidgeon, and Paul R Hunter. “Perceptions of Drinking Water Quality and Risk and Its Effect on Behaviour: A Cross-national Study.” Science of The Tota Environment 407.21 (2009):5455-468, Web. [12] Flanagan, Sara V., Steven E. Spayd, Nicholas A. Procopio, Robert G. Marvinney, Andrew E. Smith, Steven N. Chillrud, Stuart Braman, and Yan Zheng. "Arsenic in Private Well Water Part 3 of 3: Socioeconomic Vulnerability to Exposure in Maine and New Jersey." Science of The Total Environment 562 (2016): 1019-030. Web.